Measuring the Impact of Lexical Training Data Coverage on Hallucination Detection in Large Language Models

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Zhang, Shuo, Gotti, Fabrizio, Mo, Fengran, Nie, Jian-Yun
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866912724893564928
author Zhang, Shuo
Gotti, Fabrizio
Mo, Fengran
Nie, Jian-Yun
author_facet Zhang, Shuo
Gotti, Fabrizio
Mo, Fengran
Nie, Jian-Yun
contents Hallucination in large language models (LLMs) is a fundamental challenge, particularly in open-domain question answering. Prior work attempts to detect hallucination with model-internal signals such as token-level entropy or generation consistency, while the connection between pretraining data exposure and hallucination is underexplored. Existing studies show that LLMs underperform on long-tail knowledge, i.e., the accuracy of the generated answer drops for the ground-truth entities that are rare in pretraining. However, examining whether data coverage itself can serve as a detection signal is overlooked. We propose a complementary question: Does lexical training-data coverage of the question and/or generated answer provide additional signal for hallucination detection? To investigate this, we construct scalable suffix arrays over RedPajama's 1.3-trillion-token pretraining corpus to retrieve $n$-gram statistics for both prompts and model generations. We evaluate their effectiveness for hallucination detection across three QA benchmarks. Our observations show that while occurrence-based features are weak predictors when used alone, they yield modest gains when combined with log-probabilities, particularly on datasets with higher intrinsic model uncertainty. These findings suggest that lexical coverage features provide a complementary signal for hallucination detection. All code and suffix-array infrastructure are provided at https://github.com/WWWonderer/ostd.
format Preprint
id arxiv_https___arxiv_org_abs_2511_17946
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Measuring the Impact of Lexical Training Data Coverage on Hallucination Detection in Large Language Models
Zhang, Shuo
Gotti, Fabrizio
Mo, Fengran
Nie, Jian-Yun
Computation and Language
Artificial Intelligence
Hallucination in large language models (LLMs) is a fundamental challenge, particularly in open-domain question answering. Prior work attempts to detect hallucination with model-internal signals such as token-level entropy or generation consistency, while the connection between pretraining data exposure and hallucination is underexplored. Existing studies show that LLMs underperform on long-tail knowledge, i.e., the accuracy of the generated answer drops for the ground-truth entities that are rare in pretraining. However, examining whether data coverage itself can serve as a detection signal is overlooked. We propose a complementary question: Does lexical training-data coverage of the question and/or generated answer provide additional signal for hallucination detection? To investigate this, we construct scalable suffix arrays over RedPajama's 1.3-trillion-token pretraining corpus to retrieve $n$-gram statistics for both prompts and model generations. We evaluate their effectiveness for hallucination detection across three QA benchmarks. Our observations show that while occurrence-based features are weak predictors when used alone, they yield modest gains when combined with log-probabilities, particularly on datasets with higher intrinsic model uncertainty. These findings suggest that lexical coverage features provide a complementary signal for hallucination detection. All code and suffix-array infrastructure are provided at https://github.com/WWWonderer/ostd.
title Measuring the Impact of Lexical Training Data Coverage on Hallucination Detection in Large Language Models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2511.17946